The goal of this study is to evaluate the efficacy of the newly available benzodiazepine (BZ) antagonist flumazenil for rapidly detoxifying patients from BZ agonist drugs. Conventional BZ detoxification involves prescription of a long-acting BZ and gradual taper. This is problematic in outpatients being treated for substance abuse because patients often continue to supplement the prescribed benzodiazepines with illicit ones, and inpatient hospitalization is not always readily available. Flumazenil (previously known as Ro 15-1788) was recently released by the FDA in February 1992 (trade name "Mazicon") to reverse BZ overdose. The initial dose of flumazenil can induce a brief BZ withdrawal. Subsequent doses of flumazenil, however, have evoked a greatly attenuated withdrawal in studies using BZ dependent nonhuman primates, suggesting that BZ dependence and withdrawal liability had been reversed by the initial flumazenil dose. While this reversal of dependence is clearly potentially beneficial, the initial withdrawal precipitation carries some risk. Six of 446 patients receiving flumazenil in cases of suspected BZ overdose had seizures. Barbiturate pretreatment should counteract the reduction in chloride channel opening accompanying BZ receptor occupation by flumazenil in BZ dependent subjects and thus attenuate the withdrawal manifestations and reduce the seizure risk. This is analogous to the well-known clonidine reduction in the severity of opiate withdrawal precipitated by naltrexone. Each patient will participate in four sequential study periods: 1) the suitability evaluation period, 2) the stabilization period, 3) the detoxification period, and 4) the efficacy evaluation period. In the suitability evaluation period, patients of any gender or race with a structured interview (SCID) diagnosis of BZ dependence who meet inclusion and exclusion criteria will be admitted for 20 days to a clinical research inpatient bed at the Connecticut Mental Health Center. Patients will then undergo a pentobarbital challenge test in order to estimate the approximate level of daily BZ requirement. During the stabilization period, patients will initially be started on the BZ agonist alprazolam (ALP) in a daily dose as predicted from the pentobarbital challenge. The ALP dose will be adjusted as needed over the first three hospital days an then maintained at the same dosage over the next seven hospital days prior to detoxification. The 5 hour detoxification procedure will utilize a single administration of flumazenil (n=20) intravenously preceded by a single oral dose of the barbiturate anticonvulsant phenobarbital. A parallel control group will receive placebo (n=20) instead of flumazenil in double blind fashion. Whether flumazenil reverses BZ dependence will be assessed by twice daily measurement of BZ abstinence symptoms and signs in both groups after abrupt ALP discontinuation via single blind substitution of placebo over hospital days 11 to 20.